[Rd] Tricking Promises into Sending Info Via Args into Caller
The is.pos function below results in the variable, out, being set to TRUE if the first argument to is.pos is positive and to FALSE otherwise. It does this without using the return value or using scoping tricks to reach into the caller. Instead it tricks the promise into communicating one bit of information upwardly from the function to its caller via the second argument. One would have thought this to be impossible. Is this intended behavior? is.pos - function(i, x) { if (i 0) x; NULL } # in this example actual arg1 of is.pos is positive out - FALSE is.pos(1, out - TRUE) out # TRUE # in this example actual arg1 of is.pos is negative out - FALSE is.pos(-1, out - TRUE) out # FALSE -- Statistics Software Consulting GKX Group, GKX Associates Inc. tel: 1-877-GKX-GROUP email: ggrothendieck at gmail.com __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Tricking Promises into Sending Info Via Args into Caller
On Jan 12, 2013, at 17:02 , Gabor Grothendieck wrote: The is.pos function below results in the variable, out, being set to TRUE if the first argument to is.pos is positive and to FALSE otherwise. It does this without using the return value or using scoping tricks to reach into the caller. Instead it tricks the promise into communicating one bit of information upwardly from the function to its caller via the second argument. One would have thought this to be impossible. Is this intended behavior? Yes, this is a generic consequence of lazy evaluation: delayed and unpredictable side effects. Whether it is desirable is an open issue; it is the sort of thing that creates serious headaches for compiler constructors, but it is pretty much unavoidable once you include the lazy eval feature. is.pos - function(i, x) { if (i 0) x; NULL } # in this example actual arg1 of is.pos is positive out - FALSE is.pos(1, out - TRUE) out # TRUE # in this example actual arg1 of is.pos is negative out - FALSE is.pos(-1, out - TRUE) out # FALSE -- Statistics Software Consulting GKX Group, GKX Associates Inc. tel: 1-877-GKX-GROUP email: ggrothendieck at gmail.com __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel -- Peter Dalgaard, Professor, Center for Statistics, Copenhagen Business School Solbjerg Plads 3, 2000 Frederiksberg, Denmark Phone: (+45)38153501 Email: pd@cbs.dk Priv: pda...@gmail.com __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Tricking Promises into Sending Info Via Args into Caller
In the spirit of this, but AFAIK not due to lazy evaluation, here's another illustration why it's easy to mistakes when doing inline assignments: x - 0 TRUE (x - 1) [1] TRUE x [1] 1 FALSE (x - 2) [1] FALSE x [1] 1 (x - 3) FALSE [1] FALSE x [1] 3 FALSE (x - 4) [1] FALSE x [1] 4 /Henrik On Sat, Jan 12, 2013 at 9:19 AM, peter dalgaard pda...@gmail.com wrote: On Jan 12, 2013, at 17:02 , Gabor Grothendieck wrote: The is.pos function below results in the variable, out, being set to TRUE if the first argument to is.pos is positive and to FALSE otherwise. It does this without using the return value or using scoping tricks to reach into the caller. Instead it tricks the promise into communicating one bit of information upwardly from the function to its caller via the second argument. One would have thought this to be impossible. Is this intended behavior? Yes, this is a generic consequence of lazy evaluation: delayed and unpredictable side effects. Whether it is desirable is an open issue; it is the sort of thing that creates serious headaches for compiler constructors, but it is pretty much unavoidable once you include the lazy eval feature. is.pos - function(i, x) { if (i 0) x; NULL } # in this example actual arg1 of is.pos is positive out - FALSE is.pos(1, out - TRUE) out # TRUE # in this example actual arg1 of is.pos is negative out - FALSE is.pos(-1, out - TRUE) out # FALSE -- Statistics Software Consulting GKX Group, GKX Associates Inc. tel: 1-877-GKX-GROUP email: ggrothendieck at gmail.com __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel -- Peter Dalgaard, Professor, Center for Statistics, Copenhagen Business School Solbjerg Plads 3, 2000 Frederiksberg, Denmark Phone: (+45)38153501 Email: pd@cbs.dk Priv: pda...@gmail.com __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
Re: [Rd] Patch for setwd() to show path in error message
On Fri, Jan 11, 2013 at 10:40 PM, Prof Brian Ripley rip...@stats.ox.ac.uk wrote: On 11/01/2013 20:22, Henrik Bengtsson wrote: Below is a patch for setwd() to show path in error message. Current it just gives error messages such as: Error in setwd(libdir) : cannot change working directory with the patch it should (read untested) give: Error in setwd(libdir) : cannot change working directory to 'path/that/fails/' And why is this needed? You know it is 'libdir' and could check what that was (or if doing this programatically, check its existence first). Risking to state the obvious: To me, the main purpose of error messages is to make troubleshooting easier. The more informative the error message is, the easier it is to troubleshoot and the faster the origin of the error can be resolved. The other extreme is to provide an empty error message (stop()) and leave it to whoever needs to troubleshoot to use traceback() and debug() and/or inspect the source code. Also, a well written error message makes it more likely that the end user will be able to resolve the problem without asking others for help (read r-help, the maintainer, an R savvy person that the user happens to know, ...). As an example of this, over at aroma-project.org I receive fewer support emails after updating the format of file-not-found error messages from: Pathname not found: R-devel/src/library/base/R/R/utils.R to: Pathname not found: R-devel/src/library/base/R/R/utils.R (R-devel/src/library/base/R/ exists, but nothing beyond; current directory is 'x:/CRAN') Moreover, it is not always the person who receives/reads the error message that is the one who will resolve the problem. In some cases it is not even possible for that person to access the environment where the problem occurs. However, as a good R citizen, the more troubleshooting you can at your end before requesting help by the person who can take action the better. This also makes it more likely that the error will be resolved sooner. As my postscript hinted to, this post originates from an error in the 'R CMD check' service ran by r-forge.r-project.org). That particular error is likely to occur internally in tools:::.check_packages(); if (nzchar(libdir)) { setwd(libdir) ... } where 'libdir' is an internal object, but there is no way for me (and all the others who also see this error for their packages) to help out with the troubleshooting. /Henrik As you fumbled towards, it would require all the translations to be redone, so we don't do this sort of thing lightly. (Redone twice if it were done correctly, AFAICS.) The .pot and .po files are generated automatically, but not filled in automatically. This is all in the manuals In any case, the correct procedure was to file an enhancement request on R-bugs, with a case and after testing on both platforms you are proposing to patch. PATCH: % diff -u util.c util,patch.c --- util.c 2012-12-16 13:13:05.002249900 -0800 +++ util,patch.c2013-01-11 11:52:16.843236400 -0800 @@ -696,14 +696,14 @@ { const wchar_t *path = filenameToWchar(STRING_ELT(s, 0), TRUE); if(_wchdir(path) 0) - error(_(cannot change working directory)); + error(_(cannot change working directory to '%s'), path); That is a wchar_t*, so the format would need to be %ls. } #else { const char *path = R_ExpandFileName(translateChar(STRING_ELT(s, 0))); if(chdir(path) 0) - error(_(cannot change working directory)); + error(_(cannot change working directory to '%s'), path); } #endif return(wd); TRANSLATIONS: I have zero experience with GNU gettext/*.po files, but I understand that the above change requires updates to src/library/base/*.po files, which now looks like (e.g. de.po): #: src/main/util.c:699 src/main/util.c:706 msgid cannot change working directory msgstr kann Arbeitsverzeichnis nicht wechseln A generic/automatic update would be to use the format: #: src/main/util.c:699 src/main/util.c:706 msgid cannot change working directory: '%s' msgstr kann Arbeitsverzeichnis nicht wechseln: '%s' but I don't think that is in line with the current R style. /Henrik PS. Background for this patch: https://r-forge.r-project.org/tracker/index.php?func=detailaid=2395group_id=34atid=194 __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel -- Brian D. Ripley, rip...@stats.ox.ac.uk Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UKFax: +44 1865 272595 __ R-devel@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-devel